J. AMER.SOC.HORT.SCI. 137(5):325–332. 2012. Genetic Variation and Distribution of Pacific Crabapple Kanin J. Routson1 University of Arizona, Arid Lands Resource Sciences, 1955 East Sixth Street, P.O. Box 210184, Tucson, AZ 85719 Gayle M. Volk and Christopher M. Richards National Center for Genetic Resources Preservation, U.S. Department of Agriculture, 1111 S. Mason Street, Fort Collins, CO 80521 Steven E. Smith School of Natural Resources and the Environment, University of Arizona, P.O. Box 210043, Tucson, AZ 85721 Gary Paul Nabhan The Southwest Center and School of Geography and Development, University of Arizona, 1052 N. Highland Avenue, Tucson, AZ 85721 Victoria Wyllie de Echeverria School of Environmental Studies, University of Victoria, P.O. Box 3060 STN CSC, Victoria, British Columbia V8W 3R4, Canada
ADDITIONAL INDEX WORDS. species distribution modeling, genetic diversity, Malus fusca, ex situ gene conservation
ABSTRACT.Pacificcrabapple[Malus fusca (Raf.) C.K. Schneid.] is found in mesic coastal habitats in Pacific northwestern North America. It is one of four apple species native to North America. M. fusca is culturally important to First Nations of the region who value and use the fruit of this species as food, bark and leaves for medicine, and wood for making tools and in construction. However, little is known about either distribution or genetic diversity of this species. To correct this deficiency, we used habitat suitability modeling to map M. fusca habitat types with species occurrence records. The species apparently occupies at least two distinct climate regions: a colder, drier northern region and a warmer, wetter southern region. Total area of modeled habitat encompasses ’356,780 km2 of low-lying areas along the Pacific coast. A total of 239 M. fusca individuals sampled from across its native range were genetically compared using six microsatellite markers to assess for possible geographic structuring of genotypes. The primers amplified 50 alleles. Significant isolation by distance was identified across the ’2600 km (straight line) where samples were distributed. These results may help establish priorities for in situ and ex situ M. fusca conservation.
Wild apple species (Malus Mill.) are native throughout Minnesota to Texas. These have been determined to be closely temperate climes of Asia, Europe, and North America (Brown, related based on isozymes (Dickson et al., 1991). The species 2012; Luby, 2003). They offer promising sources of genetic native to the Pacific northwestern North America, M. fusca, diversity for apple breeding (Brown, 2012) and also provide is the sole geographic, morphological (Van Eseltine, 1933), a wildlife habitat and serve as a direct food source for humans. chemical (Williams, 1982), and genetic outlier among the North Four Malus species are native to North America. Three species American taxa. occur in eastern and midwestern United States and eastern Pacific crabapple is a small, deciduous, often multitrunked Canada: M. angustifolia (Aiton) Michx. is native from southern tree or thicket-forming shrub that occurs naturally in mesic New Jersey to Florida, M. coronaria (L.) Mill. from Ontario to environments along river bottoms, meadows, and muskeg South Carolina, and M. ioensis Alph. Wood Britton ranges from fringes at low to mid-elevations along the Pacific coast of North America from northern California to the Kenai Peninsula Received for publication 29 May 2012. Accepted for publication 12 July 2012. in Alaska (Viereck and Little, 1986). Yellow to red fruit are We sincerely thank all who have contributed to the success of this research. We oblong and 1 to 2 cm in length (Fig. 1). It groups genetically specifically acknowledge our British Columbia collaborators Nancy Turner, with the species native to central Asia and China rather than the Leslie Main Johnson, and Ken Downs for their insights samples from the region. other North American taxa, according to amplified fragment Thanks to Phillip Jenkins and Sarah Hunkins at the University of Arizona length polymorphism data (Qian et al., 2006) and nuclear Herbarium (ARIZ) for assistance with herbarium records and Steffi Ickert-Bond at the University of Alaska Museum of the North Herbarium (ALA) and staff at ribosomal and chloroplast DNA (Robinson et al., 2001). M. fusca the University of British Columbia Herbarium (UBC) and University of Alberta is considered to be in section Kansuenses Rehd. along with Herbarium (ALTA) for specimens. Thanks to Sarah Hayes, Joseph Postman, M. kansuensis (Batalin) C.K. Schneid., M. toringoides (Rehder) and all who helped with sample collection. Thanks to Adam Henk for technical Hughes, and M. transitoria (Batalin) C.K. Schneid. (Robinson assistance in the laboratory and Ned Garvey and Karen Williams for securing Plant Germplasm Collection funds. Finally, thanks to the Kellogg Program of et al., 2001). Because of genetic grouping with central Asia, the the University of Arizona Southwest Center for funding. species is hypothesized to be a recent migrant across the Bering 1Corresponding author. E-mail: [email protected]. Strait (Williams, 1982).
J. AMER.SOC.HORT.SCI. 137(5):325–332. 2012. 325 Materials and Methods
SAMPLE COLLECTIONS. Two hundred thirty-nine M. fusca individuals were sam- pled across the species’ native range from southeastern Alaska to northern California. Samples were obtained from field collections (138 individuals from California, Oregon, Washington, and British Columbia), herbar- ium specimens (65 individuals from Alaska and British Columbia), and U.S. Department of Agriculture (USDA) germplasm accessions (36 individuals from California, Oregon, and Washington). The latter were obtained from the USDA Agricultural Research Service Plant Genetic Resources Unit (PGRU) collec- tion in Geneva, NY. The University of Alaska Museum of the North Herbarium and Univer- sity of British Columbia Herbarium provided herbarium specimens. Field collections took place Oct. 2010. Collections were made in publicly accessible parks, natural areas, and roadsides in moist, coastal habitats. Fresh leaf tissue was sam- pled (generally five leaves per tree) for genetic analysis. The fresh leaves were stored in plastic bags and keptcooluntiltissue samples were sent to the laboratory. There they were loaded into DNA extraction plates, after which they were frozen at –80 °C for long-term storage. General physical site characteristics, habitat type, associated veg- etation, and locality descriptions were docu- mented and photographs taken at each site. Latitude/longitude coordinates and elevation were recorded for individual trees during Fig. 1. Images of Malus fusca fruit collected from trees in (A) Washington and (B) California. field collections using a handheld GPS (eTrex-Vista; Garmin, Olathe, KS). Herbar- Malus fusca can serve as a rootstock for domesticated apple ium vouchers were collected at six sites and sent to the trees in waterlogged sites (R. Duncan, personal communica- University of Arizona and University of Washington herbariums. tion) and remains a culturally important species for First Na- Locations for herbarium specimens and USDA accessions were tions of the Pacific northwestern North America (Downs, 2006; verified by comparing collection notes with Google Earth (Ver- McDonald, 2005; Turner and Turner, 2008). All parts of the tree sion 6.2; Google, Mountain View, CA) and biogeoreferencing have been used including the fruit as food, the bark and leaves web application GEOLocate Web (Rios and Bart, 2005). Points for medicines, and the dense wood for implements and in con- were selected based on habitat characteristics and descrip- struction (Turner and Bell, 1971a, 1971b). It is well docu- tions of the localities. Specimens with less than a 5000-m un- mented that Haida, Tsimshian, Tlingit, and Wakashan peoples certainty in GEOLocate were included in the species distribution in British Columbia and southeastern Alaska have tended small modeling. orchards of M. fusca that were often owned and managed in SPECIES DISTRIBUTION MODELING. A SDM for M. fusca was ways that were passed down between generations (Deur and created using MaxEnt software [Version 3.3.3 (Phillips et al., Turner, 2005; Downs, 2006; McDonald, 2005; Turner and 2006)] and WorldClim 1.4 spatially interpolated climate grids Peacock, 2005; Turner and Turner, 2008). Diploid M. fusca (Hijmans et al., 2005) and M. fusca collection localities (n = 205 hybridizes with M. ·domestica Borkh. (Hartman, 1929) and with less than 5000-m uncertainty) and M. fusca occurrence has been a source of fire blight (Erwinia amylovora Burrill) records (n = 152 with less than 5000-m uncertainty) from the resistance in apple breeding programs that use transgenic Global Biodiversity Information Facility (2001). We used early flowering methods to reduce the length of the generation 1950–2000 data at a 30-arc second resolution ( 0.56 km2 at cycles (Flachowsky et al., 2011). lat. 49° N) for elevation and 19 bioclimatic variables based on We describe genetic diversity and distribution of M. fusca monthly precipitation and temperature. American Standard through species distribution modeling (SDM) of known occur- Code for Information Interchange (ASCII) files were generated rence records and genetic fingerprinting of 239 M. fusca in DIVA-GIS [Version 7.5.0 (Hijmans et al., 2001)] from individuals sampled from northern California to southeastern WorldClim data for the Pacific northwestern North America Alaska using six microsatellite markers. region (long. 121.0° W to 151.0° W, lat. 40.0° N to 62.0° N).
326 J. AMER.SOC.HORT.SCI. 137(5):325–332. 2012. Multiple sample occurrences in grid cells (n = 149) were AP1. Unlinked primers (GD12, GD15, GD96, GD142, GD147, deleted using ENMTools (Warren et al., 2010) resulting in a and GD162) described in Hemmat et al. (2003) and Hokanson total of 208 occurrence locations to build the model. A target et al. (1998) were used to amplify microsatellite loci. Infrared group background occurrence file (Elith et al., 2011; Phillips florescent dye IRD700 or IRD800 (MWG-Biotech, High Point, et al., 2009) was constructed using 10,000 randomly selected NC) labeled forward primers. Reverse primers were unlabeled occurrence records for flowering plants from the Global Bio- (IDT, Coralville, IA). diversity Information Facility (2001). Multiple occurrences Polymerase chain reactions (PCRs) were carried out in 15-mL within a grid were removed in ENMTools. We ran MaxEnt reactions. Each reaction contained: 0.25 mL GoTaqÒ Flexi Taq using default settings. A model was selected that 1) minimized Polymerase (Promega, Madison, WI) (5 U/mL), 3 mL Promega Akaike information criterion scores [AICc (Warren and Seifert, 5· Colorless GoTaqÒ Flexi Buffer (10 mM Tris-HCl, 50 mM 2011)] calculated in ENMTools over 10 replicate runs; 2) used KCl, and 0.5% Triton X-100), 1.5 mL of 0.25 mM MgCl2, 1.5 mL environmental variables whose correlation was less than 0.60; of 0.25 mM dNTPs, and 0.25 mL forward and reverse primers. 3) showed relatively high area under curve (AUC) scores; and We added 5 mL of undiluted genomic DNA product obtained 4) appeared consistent with known species occurrences. AUC from the Qiagen DNeasy 96 plant kits to each reaction. scores relate to the probability of the model correctly scoring Addition of sterile distilled H2O brought reaction volumes to random presence and absence sites (Fielding and Bell, 1997; 15 mL. Primer pairs GD96, GD15, GD142, GD147, and GD162 Phillips et al., 2009). were multiplexed. GD12 was run independently. Malus fusca occurs across a wide latitudinal range and may PCR was run on a PTC 200 Thermocycler (MJ Research, occupy fundamentally different environments across this range Reno, NV) using touchdown PCR. The annealing temperature (Nakazato et al., 2010). We assessed environmental variability was reduced 1 °C at each cycle, beginning at 63 °C and ending across the M. fusca range by conducting a hierarchical cluster at 54 °C, followed by annealing at 55 °C for 18 cycles, and analysis using PROC CLUSTER (Ward’s minimum-variance ending with a 72 °C extension for 2 min. PCR products were method) in SAS (Version 9.3; SAS Institute, Cary, NC). Ob- diluted 1:1 with a formamide bromophenol blue loading buffer servations were based on individual occurrences and clustering and denatured at 95 °C for 5 min. The denatured products were was performed on a subset of environmental variables selected run on gels (6.5% KB Plus acrylamide; LI-COR, Lincoln, NE) to maximize variation and standardized to mean = 0 and SD = 1. in 1· TBE buffer (89 mM Tris, 89 mM boric acid, and 20 mM We selected environmental variables by first performing a EDTA) for 1 h, 45 min at 1500 V, 40 W, 40 mA, and 45 °C in hierarchical of clustering of the variables themselves using a LI-COR 4200 DNA Sequencer. Digital images of the gels PROC VARCLUS in SAS to identify correlation among vari- imported to LI-COR Saga Generation 2 software were visually ables and identify variables from non-overlapping clusters that analyzed in Saga. Overloaded gels were diluted 1:10 with maximized variation. Intrataxon clusters were identified by additional loading buffer and rerun. examining the cubic clustering criterion and the pseudo F and t2 GENETIC DATA ANALYSIS. Allele frequencies, observed het- statistics where P # 0.01 (Cooper and Milligan, 1988). Two erozygosity (Ho) and expected heterozygosity (He), allelic percent of the observations with the lowest estimated proba- richness, and Wright’s F-statistics were calculated in GDA bility density were omitted from clustering. In cases in which (Lewis and Zaykin, 2002), FSTAT (Goudet, 1995), and Genodive more than one cluster was identified, the omitted observations (Meirmans and Van Tienderen, 2004) programs. Neighbor were grouped into the nearest cluster. Separate SDMs were joining was computed in DARwin[Version5.0.158(Perrier developed for significantly distinct climate clusters. and Jacquemoud-Collet, 2006)] from dissimilarity among Suitable habitat was quantified using Arcmap (ArcGIS 10; genotypes (Perrier et al., 2003). Both dissimilarity calculations Esri, Redlands, CA). Probability of presence/background ASCII and neighbor joining were bootstrapped over 10,000 replica- files produced by MaxEnt were converted into raster files in tions. Isolation by distance was determined by a Mantel (1967) Arcmap. Map projections were converted from WGS1984 into test using Rousset’s linearized F-statistic (F ), FST (Rousset, ST 1 FST Albers equal-area conic projection to calculate area of suitable 1997), over 5000 permutations in Genodive. À habitat. The percentage of suitable habitat sampled in the genetic We used STRUCTURE software (Pritchard et al., 2000) to analyses was calculated by applying a 10-km buffer to each identify possible evidence of population structure using poste- individual included in the genetic analyses, then merging rior probabilities of possible populations. The number of pop- buffered areas and clipping to exclude non-suitable habitat ulations was estimated from the natural log of the probability of using the Buffer and Clip functions in Arcmap. allele frequencies over the posterior probability of one to 12 MICROSATELLITE ANALYSIS. The genetic analysis of the 239 possible populations (Evanno et al., 2005). An ancestry model samples was performed following procedures described in Volk of admixture and correlated allele frequencies between pop- et al. (2005). The six unlinked microsatellite markers used in ulations enabled fractional assignment of individual genotypes this study can only represent 35% of the 17 linkage groups in to multiple populations by probability of membership. We ran the Malus genome. However, they have been used to success- a Markov chain Monte Carlo method in STRUCTURE using an fully differentiate between individuals and identify structure in iteration burn-in period of 500,0000 runs followed by 100,000 other Malus sp. (Richards et al., 2009; Volk et al., 2008). We iterations per chain. We ran each Markov chain 100 times for extracted genomic DNA from frozen leaf tissue (field collec- one to 12 possible populations. STRUCTURE output was tions) and dried leaf tissue (herbarium records) using DNeasy captured and analyzed by STRUCTURE HARVESTER (Earl 96 plant kits (Qiagen, Valencia, CA). DNA was extracted from and vonHoldt, 2011). CLUMPP software produced population the frozen leaves using 50 mg tissue frozen in liquid nitrogen membership coefficients based on matrices of multiple during initial tissue grinding; 10 to 12 mg tissue was used for STRUCTURE runs to average multiple population assignment the dry samples and was extracted using the same protocol runs for individuals belonging to more than one cluster but without liquid nitrogen and a lower temperature of reagent (Jakobsson and Rosenberg, 2007). Individual genotypes were
J. AMER.SOC.HORT.SCI. 137(5):325–332. 2012. 327 assigned to the population cluster in which they had the highest Potential M. fusca habitat was mapped using two SDMs for membership. the northern and southern climatic clusters (Fig. 2). The SDM for the northern cluster for M. fusca was developed using eight Results and Discussion WorldClim bioclimatic variables. Bioclimatic variables of altitude, mean diurnal temperature range, minimum tempera- SPECIES DISTRIBUTION MODELING. Cluster analysis of six ture of coldest month, mean temperature of driest quarter, mean bioclimatic variables on 157 georeferenced individuals identi- temperature of warmest quarter, annual precipitation, precipi- fied two significant climatic clusters (Fig. 2), revealing a colder tation of driest month, and precipitation of coldest quarter were and drier ‘‘northern’’ cluster with 47 individuals (mean annual included as model elements. The SDM for the southern climatic temperature = 4.2 °C, annual precipitation = 1702 mm) and cluster used seven bioclimatic variables including: altitude, a warmer and wetter ‘‘southern’’ cluster with 110 individuals annual mean temperature, mean diurnal temperature range, (mean annual temperature = 9.1 °C, annual precipitation = mean temperature of wettest quarter, annual precipitation, 2267 mm). Bioclimatic variables of temperature seasonality, precipitation of warmest quarter, and precipitation of coldest mean temperature of wettest quarter, mean temperature of quarter as model elements. The models were selected based warmest quarter, annual precipitation, precipitation of driest on low AICc scores [6134 (northern), 5708 (southern)] calcu- month, and precipitation of coldest quarter were used in the lated over 10 replicate runs, less than 0.60 correlation among cluster analyses. These variables were selected from the PROC environmental variables, high AUC scores (AUC = 0.985 in the VARCLUS hierarchical of clustering on the variables them- northern cluster, AUC = 0.986 in the southern cluster), and selves to identify variables from non-overlapping clusters that consistency with known species occurrences. When the two maximized variation (see ‘‘Materials and Methods’’). The two SDMs were merged in ArcMap, there is 356,780 km2 of climatic clusters are not perfectly segregated in space. Two potential habitat for M. fusca (Fig. 2). Using a buffer area of species occurrences on the north end of Haida Gwaii, British 10 km, samples representing 15,306 km2 of potential habitat Columbia, Canada, were characterized as being climatically were used in the genetic analyses, equating to 4.29% of the grouped with the ‘‘southern’’ cluster, although they are gener- potential habitat. ally well within the ‘‘northern’’ climatic region. Ocean currents GENETIC DATA ANALYSIS. The six primers amplified a total of or some ‘‘island’’ phenomenon associated with Haida Gwaii 50 alleles (Table 1). Polymorphic alleles observed in the 239 could be responsible for the mixing of clusters. individuals ranged from a low of two for GD15 to a maximum of 20 for GD147. Observed heterozygosity (0.343) was found to be lower than expected heterozygosity (0.428) suggesting slight inbreeding may be occurring (Table 1). In another apple species, M. sieversii (Ledeb.) M. Roem., Richards et al. (2009) report much higher values of observed and expected heterozy- gosities (He = 0.749; Ho = 0.693). This may indicate higher total genetic diversity in M. sieversii compared with M. fusca. The six primers were sufficient to differentiate among all individuals except nine pairs of genotypes. Seven of these were from adjacent trees and two from herbarium records. Two duplicate genotypes were found in the University of Alaska herbarium specimens, one from the vicinity of Katalia, AK, and from the Yakutat Foreland, AK, which are 300 km apart. Another herbarium specimen’s genotype from the Alexander Archipelago, AK, matched a genotype from the Skeena River, near Kitwanga, British Columbia, Canada, also more than 300 km distant. An additional seven markers, described by Liebhard et al. (2002), were run on the nine pairs of duplicates and 27 indi- viduals randomly selected from the northern (15 individuals) and southern clusters (12 individuals) based on percentage of individuals sampled from unique grid cells in each habitat to confirm the results of the six markers (Table 1). These addi- tional markers resolved the geographically isolated duplicates into unique genotypes. Two of the duplicates from adjacent trees were not resolved and are presumed to have been collected from single individuals (the shrub nature of M. fusca, possible suckering, and often densely vegetated habitat make differen- tiating between individual trees difficult in some cases). When individuals collected within 1 km of each other were assumed to be from the same population and grouped together, Fig. 2. A habitat suitability model derived from presence only data in MaxEnt we found significant differences between populations using indicates predicted suitable habitat for Malus fusca in the shaded areas. Northern (‘‘x’’ signs) and Southern (‘‘+’’ signs) denote two significantly analysis of molecular variance (Excoffier et al., 1992; Michalakis different habitat types derived from clustering of six WorldClim bioclimatic and Excoffier, 1996) with an FST = 0.074, P = 0.001 (range, 0.05 variables for M. fusca presence of data based on collection localities. to 0.099). However, there is a much higher within-population
328 J. AMER.SOC.HORT.SCI. 137(5):325–332. 2012. variability (F-statistic FIS = 0.135, P = 0.001). Overall, these This may be a product of the high admixture observed in the values are similar to those reported for other species (Table 2), species. Another possibility is that habitat-related selection or except that fewer polymorphic alleles were identified in M. fusca a previous founder event drove the species toward uniformity. It than either M. sieversii (Richards et al., 2009) or M. orientalis is also possible that not enough of the genome was sampled or Uglitzk. (Volk et al., 2008). This may indicate a lower total too few individuals were sampled to accurately capture IBD genetic variation in M. fusca compared with other wild Malus structure. M. fusca is an animal-pollinated and dispersed species. species with the potential for effective long-distance dispersal. A neighbor-joining dendrogram (Fig. 3) of the 47 individ- This, in combination with a relatively continuous current uals from the northern climatic region and 47 individuals distribution of suitable habitat (Fig. 2), lends support to the randomly selected individuals from the southern climatic genetic findings of a single continuous population with low IBD cluster show no distinct separation of clusters by region. This across the species range. suggests high admixture among M. fusca and a general lack of The 239 individuals were grouped into ‘‘population’’ clus- genetic structure. ters using posterior probabilities of allele frequencies in We found low but significant isolation by distance (IBD) for STRUCTURE software (Pritchard et al., 2000). Structure FST 1 FST by kilometers between populations (defined as all in- results identified four significant clusters. The composition of dividualsÀ within 1 km of each other). Mantel’s r was calculated individuals within clusters is highly admixed both genetically at 0.395 (P = 0.007, error sum of squares = 0.483, R2 = 0.156). and geographically. Each individual was assigned to multiple Isolation by distance reported in other species of woody clusters (the highest percentage of any individual belonging to Rosaceae range from marginal isolation by distance (R2 = a single cluster was 41% with an average highest contribution 0.17, P =0.091)across 10 km in Prunus mahaleb L. of 33% across all samples). Clusters showed no geographic populations in southeastern Spain (Jordano and Godoy, 2000) separation when individuals were plotted on Google Earth to high isolation by distance (R2 = 0.67, P = 0.01) across 2.5 km (Version 6.2; Google). in island populations of wild flowering cherry [Prunus lannesiana Pleistocene ice sheets in the Pacific northwestern North (Carriere) E.H. Wilson] from the Izu Islands in Japan (Kato America [30,000 to 10,000 C years BP (uncalibrated radiocar- et al., 2011). Grouping of samples in an IBD plot indicates bon date); Clague and James, 2002] resulted in strong geo- geographic barriers to gene flow (Guillot, 2009). The IBD plot graphic structuring of many species now found in the region for M. fusca (Fig. 4) does not show distinct groupings between (reviewed in Shafer et al., 2010). Although this may have in- samples, suggesting a single, continuous population under IBD. fluenced the genetic structure of M. fusca, either through the Although significant, these values of IBD seem very low com- extirpation of some isolated populations or through recolonization pared with the magnitude of the distance surveyed (2600 km). from isolated refugia, the high admixture and lack of discernible
Table 1. Descriptive statistics by locus.z Locus Individuals (no.) Polymorphic alleles (no.) He Ho LG GD12 237/27 5/5 0.538/0.567 0.544/0.481 3 GD15 237/27 2/2 0.159/0.307 0.127/0.296 Unknown GD96 237/27 7/6 0.311/0.361 0.309/0.370 17 GD142 236/27 10/2 0.122/0.037 0.122/0.037 9 GD147 236/27 20/13 0.859/0.866 0.410/0.556 13 GD162 239/27 6/5 0.580/0.633 0.544/0.444 4 CH01d09 27 16 0.936 0.741 12 CH02b12 27 9 0.846 0.852 5 CH02d08 26 5 0.545 0.577 11 CH05e03 21 24 0.962 0.905 2 NH009b 27 4 0.640 0.370 13 NH015a 22 15 0.933 0.636 Unknown NZ28f4 27 6 0.676 0.593 12 Mean 237/26.1 10.2/8.6 0.428/0.639 0.343/0.528 — Total 239/27 50/112 — — — zSix markers were run across 239 individuals, and 13 markers were run on 27 randomly selected individuals of Malus fusca. Reported values include number of individuals, number of polymorphic alleles, expected heterozygosity (He), observed heterozygosity (Ho), and mapped linkage group (LG) for each marker (Celton et al., 2009).
Table 2. Diversity comparisons with other Malus sp. Geographic range (maximum distance), number of markers in the analyses, number of amplified alleles, within population (pop.) variance (FIS), and among population variance (FST) are presented for M. fusca, M. sieversii (Richards et al., 2009), and M. orientalis (Volk et al., 2008).
Species Geographic range (km) Markers (no.) Alleles (no.) FIS FST Reference M. fusca 2600 6 50 0.135 0.074 Present study M. sieversii 1100 7 103 0.190 0.050 Richards et al., 2009 M. orientalis 760 7 115 0.208 0.046 Volk et al., 2008
J. AMER.SOC.HORT.SCI. 137(5):325–332. 2012. 329 including effective dispersal mechanisms or movement of plant material by humans or animals. The USDA PGRU conserves M. fusca germplasm ex situ as 40 living accessions obtained from California, Oregon, Washing- ton, and one from Alaska. Thirty-four of the 50 alleles found in the field collections and herbarium specimens are present in the USDA accessions. An additional 11 alleles present in the USDA accessions were not identified in field collections or herbarium records. These additional alleles could be present in wild populations not sampled or could be hybrid introgressions. FST between USDA collections and the herbarium speci- mens and the field collections showed slight, significant differences between the sample sets (FST = 0.006, P = 0.043), which suggest that USDA PGRU M. fusca collections gen- erally represent wild populations, at least in the regions sampled. However, the USDA collection contains only a single representa- Fig. 3. Interindividual distances among Malus fusca genotypes. Individuals are labeled by northern (squares) and southern (circles) climatic region where each individual was collected. The lack of tive sample from the northern climatic clus- distinct grouping between climatic region by cluster signifies high admixture in M. fusca. ter, and although large differences in allele frequencies between the climatic clusters were not identified in the neutral markers used in this study, a comprehensive collec- tion effort of the species may warrant addi- tional collections from the northern climatic cluster.
Conclusions
We estimate suitable M. fusca habitat to be geographically limited to 356,780 km2. The habitats within this range consist of low- lying, moist areas along the northern Pacific coast. In the regions sampled (less than 5% of suitable habitat), M. fusca shows high ad- mixture with little population differentiation, although it does show significant IBD across the 2600 km (straight line) sampled. High admixture in M. fusca could be a result of a continuous habitat distribution, effective pollen and seed dispersal mechanism, recent introduction and rapid dispersal of the spe- cies, trade or movement of M. fusca plant material by humans, or a combination of factors. This research can be brought to bear Fig. 4. Isolation by distance regression. Malus fusca shows significant isolation by distance as on ex situ and in situ conservation manage- demonstrated in this biplot displaying geographical distance between sampling sites and genetic ment of the species. USDA germplasm col- distance using Rousset’s distance measure (Rousset, 1997). Individuals collected 1 km or less apart lection currently contains only a single were grouped into populations. Fst is a measure of genetic differentiation between groups of gentoypes (R2 = 0.162, P = 0.0019). representative sample of M. fusca from the northern regions of its range. Differences in the climate regions could justify additional collections in the northern region, although geographic patterning in STRUCTURE results suggest a the PGRU does capture much of the identified variability. Much lack of geographic structuring of M. fusca populations, of the M. fusca range covers regions not readily threatened by lending support to a recent introduction of the species in immediate coastal development or urban expansion. On the the last 10,000 years. Other mechanisms could also account other hand, warmer temperatures and higher summer moisture for a lack of geographic structuring and high admixture deficits predicted under climate model simulations (Littell
330 J. AMER.SOC.HORT.SCI. 137(5):325–332. 2012. et al., 2010; Mote and Salath´e, 2010) could negatively affect genetic diversity and relationships in a Malus · domestica Borkh. M. fusca distribution in the future. core collection. Theor. Appl. Genet. 97:671–683. Jakobsson, M. and N.A. Rosenberg. 2007. CLUMPP: A cluster Literature Cited matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics Brown, S.K. 2012. Apple, p. 329–367. In: Badenes, M.L. and D.H. 23:1801–1806. Byrne (eds.). Fruit breeding. Springer, New York, NY. Jordano, P. and J.A. Godoy. 2000. RAPD variation and population Celton, J.M., D. Tustin, D. Chagn´e, and S. Gardiner. 2009. Construc- genetic structure in Prunus mahaleb (Rosaceae), an animal-dispersed tion of a dense genetic linkage map for apple rootstocks using SSRs tree. Mol. Ecol. 9:1293–1305. developed from Malus ESTs and Pyrus genomic sequences. Tree Kato, S., Y. Tsumura, H. Iwata, and Y. Mukai. 2011. Genetic structure Genet. Genomes 5:93–107. of island populations of Prunus lannesiana var. speciosa revealed by Clague, J.J. and T.S. James. 2002. History and isostatic effects of the chloroplast DNA, AFLP and nuclear SSR loci analyses. J. Plant Res. last ice sheet in southern British Columbia. Quat. Sci. Rev. 21:71–87. 124:11–23. Cooper, M.C. and G.W. Milligan. 1988. The effect of measurement Lewis, P.O. and D. Zaykin. 2002. GDA user’s manual. 12 July 2012. error on determining the number of clusters in cluster analysis,
J. AMER.SOC.HORT.SCI. 137(5):325–332. 2012. 331 Robinson, J.P., S.A. Harris, and B.E. Juniper. 2001. Taxonomy of the Van Eseltine, G.P. 1933. Notes on the species of apples. I. The genus Malus Mill. (Rosaceae) with emphasis on the cultivated apple, American crabapples. Agr. Expt. Sta. New York Tech. Bul. No. 208. Malus domestica Borkh. Plant Syst. Evol. 226:35–58. Viereck, L.A. and E.L. Little. 1986. Oregon crab apple (Malus fusca), Rousset, F. 1997. Genetic differentiation and estimation of gene flow p. 205–207. In: Alaska trees and shrubs. University of Alaska Press, from F-statistics under isolation by distance. Genetics 145:1219– Fairbanks, AK. 1228. Volk, G.M., C.M. Richards, A.A. Reilley, A.D. Henk, P.L. Forsline, Shafer, A.B.A., C.I. Cullingham, D.W. Coltman, and S.D. Cote. 2010. and H.S. Aldwinckle. 2005. Ex situ conservation of vegetatively Of glaciers and refugia: A decade of study sheds new light on the propagated species: Development of a seed-based core collection for phylogeography of northwestern North America. Mol. Ecol. Malus sieversii. J. Amer. Soc. Hort. Sci. 130:203–210. 19:4589–4621. Volk, G.M., C.M. Richards, A.A. Reilley, A.D. Henk, P.A. Reeves, Turner, N.C. and M.A.M. Bell. 1971a. The ethnobotany of the Coast P.L. Forsline, and H.S. Aldwinckle. 2008. Genetic diversity of wild Salish Indians of Vancouver Island. Econ. Bot. 25:63–104. Malus orientalis from Turkey and southern Russia. J. Amer. Soc. Turner, N.C. and M.A.M. Bell. 1971b. The ethnobotany of the Coast Hort. Sci. 133:383–389. Salish Indians of Vancouver Island. Appendix I. Econ. Bot. 25:335– Warren, D.L., R.E. Glor, and M. Turelli. 2010. ENMTools: A toolbox 339. for comparative studies of species distribution models. Ecogeogra- Turner, N.J. and S. Peacock. 2005. Solving the perennial paradox: phy 33:607–611. Ethnobotanical evidence for plant resource management on the Warren, D.L. and S.N. Seifert. 2011. Species distribution modeling in northwest coast, p. 101–150. In: Deur, D. and N.J. Turner (eds.). Maxent: The importance of model complexity and the performance Keeping it living. University of Washington Press, Seattle, WA. of model selection criteria. Ecol. Appl. 21:335–342. Turner, N.J. and K.L. Turner. 2008. ‘Where our women used to get the Williams, A.H. 1982. Chemical evidence from the flavonoids food’: Cumulative effects and loss of ethnobotanical knowledge and relevant to the classification of Malus species. Bot. J. Linn. Soc. practice; case study from coastal British Columbia. Botany 86:103–115. 84:31–39.
332 J. AMER.SOC.HORT.SCI. 137(5):325–332. 2012.